Abstract: The similarity measure between users has significant impact on the results of collaborative filtering recommendation system. To increase the accuracy of neighbor selection, a weighted Pearson Correlation Coefficient（PCC） similarity measurement is proposed to calculate PCC weighting factor directly with the number of user-item ratings. The improved pearson similarity metrics is applied to empirical analysis of the MovieLens, Douban and Epinions dataset. Experimental results show that the proposed method can improve the recommendation accuracy of collaborative filtering effectively in terms of Mean Absolute Error（MAE） and precision.

Key words: collaborative filtering, similarity, pearson correlation coefficient；Mean Absolute Error（MAE）

摘要： 协同过滤算法中用户相似性度量的准确性对推荐质量有显著影响。为了提高用户协同过滤算法中近邻选择的准确率，提出一种加权的皮尔逊相关系数（PCC），可根据用户-项目的评分数，直接计算出PCC加权因子。将改进的皮尔逊相似度机制用于MovieLens，Douban和Epinions数据集进行实证分析。结果表明，提出的算法可以有效提高协同过滤推荐的平均绝对误差（MAE）和准确度。

关键词: 协同过滤, 相似性, 皮尔逊相关系数, 平均绝对误差（MAE）